Nvidia Driver needs a reinstall after each reboot

H-Radical · 09-27-2007, 04:10 AM

Hi all,

I have had a serious problem since leaving edgy for feisty fawn. Every time I restart the computer, I have to reinstall the nvidia driver for it to work properly (using NVIDIA GeForce4 420 Go w/ driver 9639 on VAIO PCG-GRT100). Here's the detailed description:
-Once I install the driver, and restart X, everything is fine, and I get desktop effects working, too.
-After one reinstall (since the time X was working), I get the nvidia splash screen and the graphical login prompt, but after I log in, the gnome splash screen won't go away for about a minute (unless I click on it), and some items on the panel on the top right side won't show for about a minute, either. Desktop effects is completely gone, eventhough System > preferences > desktop effects would tell you otherwise. If I "disable" desktop effects and try to re-enable it, I get an error saying it can't be done. When I go to System > Admin > Restricted Drivers Manager I get a message saying I need to install the restricted kernel modules, eventhough I had them installed before the last reboot. Finally,

Code:

# nvidia-settings

returns a segmentation fault.
-If I reboot a second time, I get three splashed of the nvidia logo, before X completely crashes and burns and I find myself at the command prompt, reinstalling the driver and the whole story repeats itself.

I have had this problem for a long time now, and I have searched all over the place and tried all sorts of solutions that seem to have worked for others with similar problems, but they have made no difference at all for me. For example, I removed my old kernel images, and in my /etc/default/linux-restricted-modules-common I have the line:

Code:

DISABLED_MODULES = "nv"

One thing I noticed was that most people with a similar problem had to reinstall the driver after just one reboot.
The only thing I think might work and have yet to try is from this post: http://www.linuxquestions.org/questi...d.php?t=282872 but the instructions are for FC3 not Feisty Fawn and I don't have the /etc/udev/devices directory.
I am begging you, if you know anything that might help resolve this issue, please tell me. This is a laptop I'm talking about so having to reinstall a driver every second reboot is quite an annoyance.

Thanks in advance for your help.

jschiwal · 09-27-2007, 04:52 AM

You could try running a couple commands as root:
depmod -a
ldconfig

The first produces a file listing the dependencies of kernel modules. The second produces a list of libraries available in the paths listed in /etc/ld.conf.so. Also read the output of "dmesg" and the xorg log in /var/logs. If you have a video config program for you distro you might try running it. Also the NVidia installation includes /usr/bin/nvidia-xconfig that might work to config your xorg.conf file properly.

I don't have an /etc/udev/devices/ directory either, but I grep'ed in /etc/udev/udev.d/ for nvidia:

Code:

# video/graphics
SUBSYSTEM=="video4linux",       GROUP="video"
SUBSYSTEM=="graphics"           GROUP="video"
KERNEL=="vtx*|vbi*",            GROUP="video"
KERNEL=="winradio*",            GROUP="video"
KERNEL=="vttuner*",             GROUP="video"
KERNEL=="nvidia*|nvidiactl*",   GROUP="video"
KERNEL=="video0",               SYMLINK+="video"
KERNEL=="radio0",               SYMLINK+="radio"
KERNEL=="vbi0",                 SYMLINK+="vbi"
KERNEL=="pmu",                  GROUP="video"

Electro · 09-27-2007, 05:03 AM

I recommend remove the deb file for nVidia. Change the driver in xorg.conf to nv. Download the nVidia module (driver) in a tgz or run file. Install the kernel headers, compiler, and any require headers. Then backup X11 libraries. Change the runlevel to 3 or mult-user level with out GUI. As root type modprobe -r nvidia. Run the nVidia installer, but force it to compile the module. As root type modprobe -v nvidia. In xorg.conf change the driver to nvidia. Change runlevel to 5 or GUI mode. You should see nVidia splash if you have not add nologo option to xorg.conf.

In summary:
1) Remove nVidia and the nVidia deb file
2) Download nVidia run file
3) Backup X11 libraries
4) Install necessary headers and compiler for the nvidia module to compile.
5) Change runlevel to 3 or multi-user level with out GUI.
6) As root, type modprobe -r nvidia
7) Run nvidia installer and force it to compile the nvidia module
8) type modprobe -v nvidia
9) Check the version nvidia module in procfs
10 change runlevel to 5 or GUI mode
11) If nologo is set to default, should see the nVidia splash screen.

The number steps may not relate to one in the paragraph.

I have installed the nVidia module (driver) several times with the above steps. Since I am using Gentoo and the Gentoo developers provided a better way than the steps above, installing nVidia module is a lot easier than nVidia's installer.

IMHO, Gentoo is the only distribution to use if want to do the Linux way instead of some proprietary way. The Linux way is compiling the features that you want instead what the maintainer thinks that you want.

H-Radical · 10-01-2007, 11:55 PM

Hi all, and I apologize for the late response. I was away for a few days.
Thank you to jschiwal and Electro for your help. While I tried your solutions (they did not work), I accidentally did stumble upon the source of the problem and the solution. It turns out that there was a single nvidia script (/etc/rc2.d/S20nvidia-glx) that was the root cause of all this headache. Simply changing its permissions to not allow it to be executed fixed the problem. Hopefully this post will show up on the google search of the next guy who's having to reinstall the driver after each reboot.

Thanks again,
H-Radical

jlinkels · 10-26-2008, 09:24 PM

The solution worked for me. In Debian, both nvidia-glx and nvidia-kernel were not installed, BUT the /etc/init.d/nvidia-glx and /etc/init.d/nvidia-kernel script were still installed and being executed at start-up.

Removing them the Debian way

Code:

update-rc.d -f nvida-glx remove
update-rc.d -f nvida-kernel remove

worked.

After a reboot the nvidia driver loaded, glxgears shows the same performance.

jlinkels

farslayer · 10-26-2008, 10:07 PM

The simple resolution you were looking for, was actually to purge the nvidia-glx package first, then re-install the driver using the nvidia script. That would have removed that entry automatically

sudo apt-get remove --purge nvidia-glx

http://www.nvnews.net/vbulletin/showthread.php?t=72490

Quote:

Debian GNU/Linux or [K]Ubuntu with Xorg 7.x

If you wish to install the NVIDIA Linux graphics driver on a Debian GNU/Linux or Ubuntu system that ships with Xorg 7.x, please ensure that your system meets the following requirements:

* development tools like make and gcc are installed
* the linux-headers package matching the installed Linux kernel is installed
* the pkg-config and xserver-xorg-dev packages are installed
* the nvidia-glx package has been uninstalled with the --purge option and the files /etc/init.d/nvidia-glx and /etc/init.d/nvidia-kernel do not exist

If you use Ubuntu, please also ensure that the linux-restricted-modules or linux-restricted-modules-common packages have been uninstalled. Alternatively, you can edit the /etc/default/linux-restricted-modules or /etc/default/linux-restricted-modules-common configuration file and disable the NVIDIA linux-restricted kernel modules (nvidia, nvidia_legacy) via:

DISABLED_MODULES="nv nvidia_new"

Additionally, delete the following file if it exists:

/lib/linux-restricted-modules/.nvidia_new_installed

jlinkels · 10-27-2008, 04:18 AM

Hehe... eventually I forgot to paste that link in my post, it is exactly the page where I have been and found the solution.

However, nvidia-kernel and nvidia-glx were not installed. Not sure if you can remove --purge packages which are not there.

jlinkels